A R E N C I T e c h n i c a l R e p o r t

Adapting Scienti�c Work�ows onNetworked Clouds

Using Proactive Introspection

CMDS: A Data System for a Multi-institutional Mammography Registry

TR-17-02

Oleg Kapeljushink & Charles Schmitt, PhDRenaissance Computing Institute (RENCI)University of North Carolina at Chapel Hill

Louise M. Henderson, PhD, MSPH & Mikael Anne Greenwood-Hickman, MPH Department of Radiology, School of MedicineUniversity of North Carolina at Chapel Hill

Corresponding author: Louise M. Henderson, Department of Radiology , 2006 Old Clinic, CB #7150, Chapel Hill, NC, 27599, USA louise_henderson@med.unc.edu; 919.966.0525

w w w . r e n c i . o r g

Abstract

Background: Radiologists are subject tonumerous regulations, including those set forth bythe Mammography Quality Standards Act(MQSA) and the Breast Density andMammography Reporting Act (BDMRA).Compliance with these regulations can beextremely challenging, particularly for community-based and rural practices with limited budgets tosupport the adoption of a commercialMammography Information System (MIS) ormammography-supported Electronic MedicalRecord (EMR) system that can assist withregulatory requirements for patient follow-up andreporting. Since its inception in 1994, the CarolinaMammography Registry (CMR) has capturedprospective data on ~650,000 female patients and>2.4 million visits to more than 40 radiologypractices located in 39 counties in North Carolina.One of the goals of the CMR is to assistparticipating radiology practices with regulatoryrequirements. The Carolina Mammography DataSystem (CMDS) captures CMR data from eightpractices without access to a commercial MIS ormammography-supported EMR system.Objective: To describe our experience in theresearch, development, and evaluation of a majorupdate to CMDS, promoting it to a modernmedical informatics platform.Methods: We engaged in an extensiverequirements gathering phase, followed by an

iterative testing, evaluation, and modificationperiod after development of a protocol CMDS solution. A second round of iterative testing,evaluation, and modification followed “live”deployment of CMDS v1.0. After applying fixes,CMDS v1.1 was deployed and remains in use. Thesystem undergoes continuous evaluation andmodification in response to user feedback.Results: We designed the CMDS to: (1) assistparticipating CMR radiology practices in meetingthe reporting requirements of MQSA and BDMRAthrough automated generation of letters andreports; (2) provide a data repository forparticipating radiology practices and a datacollection tool for researchers interested inpopulation-level research on breast imaging andclinical outcomes; and (3) improve patient care byfacilitating the tracking and follow-up of patientsand promoting innovative outcomes-based research.Conclusion: The CMDS was custom-made tosupport the CMR, but it can be adapted for avariety of uses in radiology practice. We arecurrently modifying the system to support a LungCancer Screening Registry.Keywords: mammography; breast imaging;radiology; information systems; registries;legislation; mandatory reporting; MQSA; BDMRA

1

I. Introduction

Mammography is one of the most common imagingprocedures conducted by radiologists. Indeed, whileclinical guidelines on mammography remain a topicof great debate [1], the procedure remains theprimary means of early detection of breast cancerand reduction of associated risk of morbidity andmortality [2].

As with most healthcare professionals, radiologistsare facing increasing federal regulations regardingthe use of mammography [3,4]; these includeregulations set forth by the Mammography QualityStandards Act (MQSA) and the Breast Densityand Mammography Reporting Act (BDMRA).Compliance with these regulations can beextremely challenging, given the amount of patientfollow-up and reporting that is required. Forcommunity-based and rural radiology practiceswithout access to a Mammography InformationSystem (MIS), a mammography-supportedElectronic Medical Record (EMR) system, oranother advanced Healthcare InformationTechnology, compliance with federal regulationscan present a significant administrative andfinancial burden [5].

The Carolina Mammography Registry (CMR) wasfounded at our institution in 1994 and currentlycomprises data on ~650,000 female patients and>2.4 million visits to more than 40 radiologypractices located in 39 counties in North Carolina.One of the goals of the CMR is to facilitatecompliance with MQSA and BDMRA.The CarolinaMammography Data System (CMDS) wasdeveloped to support participation in the CMRamong practices without access to a MIS ormammography-supported EMR system.

This paper describes our experience with research,development, implementation, and evaluation ofthe CMDS. Our specific aims were to: (1)transition historical CMDS data from an Access-based data system to a new SQL-based datasystem; (2) update the CMDS to enable easymodification and continual alignment with clinicalbest practices and regulatory requirements; and (3)assist in compliance with MQSA and BDMRA.

II. MethodsAssessment of Existing SystemsPrior to development of the custom CMDS, theCMR technical and clinical/research teams engagedin a multi-month requirement gathering andevaluation period, which took place from May 2013through November 2013 and involved in-personworking sessions, phone calls, and email exchanges.The legacy CMR data system was implementedusing an Access database. However, over time, asthe CMR grew and as technologies evolved, theCMR team identified several factors that promptedthe decision to consider a new custom system,including: (1) inadequacy of Access to handle agrowing number of data fields (Access sets amaximum of 255 fields); (2) incompatibility of theold system with newer versions of Windows; (3)need for multiple users to be able to access thesystem at any given time (Access supports onlyone user at a time); and (4) requirement for anenterprise-level system to assist in compliance withfederal mandates as outlined in MQSA andBDMRA.

A variety of commercial MIS are available [6]. These include: MagView (MagView, http://www.magview.com/home), which was originally developed for the American College of Radiology; Opal-wRIS (Viztek, http://viztek.net/products/products_opal-wris/); PenView (PenRad, http://www.penrad.com/products_penview.html); RadNet® Mammography Management (Cerner Corporation, https://store.cerner.com/items/2265);Siemens Mammography Systems (Siemens Medical Solutions USA, Inc., http://usa.healthcare .siemens.com/mammography); and the MRS Suite of Systems (MRS, http://www.mrsys.co m/about). In addition, commercial EMR systems often include MIS capabilities as an add-on feature. The CMR technical and clinical/research teams evaluated existing options but concluded that none of them were sufficiently modifiable to meet the specific needs of the CMR, including the inclusion of project-specific data elements. The commercial systems also were too expensive for use in community-based and rural radiology clinics. Thus,

2

a decision was made to replace the legacy Access-based data system with a custom-made, SQL-basedsystem: the CMDS.

Development of New SystemThe CMR technical and clinical/research teammembers worked closely with one another to designthe new data system, map variables from the oldsystem to the new system, and migrate data andoperations. System development began inNovember 2013. The underlying design of the newsystem was intended to be largely generic in orderto promote reuse of design features and softwarecode for related projects (e.g., a Lung CancerScreening Project that is funded by the NationalCancer Institute). The Microsoft .NET 4.5framework was chosen as it supports the Windows7+ platforms in place at the participatingradiology practices.

The SQL database was developed after extensivemodification of the Access database, in terms ofboth layout and organization (variables andschema). Database variables, as defined in the oldschema, were manually mapped one-to-one tovariables used in the new schema. More than twohundred new variables were added to the newschema in order to improve the granularity of thedata system and better reflect current clinicalpractice and the variables in the CMR datacollection form.

Templates for follow-up letters and reports weremodified as part of the mapping and updatingprocesses and also are continuously evaluated andmodified to align with evolving federal regulationsand radiology practice needs. Finally, a new UserInterface (UI) was created to reflect the newvariables, data collection form, and templates.

Implementation of New SystemRollout of a protocol CMDS took place duringAugust 2014 and September 2014. An iterativetesting process accompanied rollout of the protocolsolution (see below). Fixes were added to thesystem, and “live” rollout of CMDS v1.0 began inJanuary 2015. Data migration of historical datafrom the Access-based system to the new SQL-based system took place practice-by-practice, either

remotely or by manual, on-site copy afterdeployment of CMDS v1.0. A second round ofiterative testing and fixes took place afterdeployment of CMDS v1.0, and “live” rollout ofCMDS v1.1 began in August 2015. CMDS v1.1remains in use today. Ongoing feedback fromparticipating radiology practices allows forcontinual refinement of the system as modificationsand fixes are compiled for future versions.

Most of the historical data were successfullymigrated to the new system. Of 616 variables inthe old schema, 108 (17.5%) were not mapped tothe new schema because of broken foreign keys orupon request of the CMR clinical/research team(i.e., they were deemed outdated or otherwiseirrelevant due to the addition of new variables).Approximately 99% of the historical data werecaptured in the new system. The CMR teammaintains an archived copy of the completehistorical data set for each participating radiologypractice.

Testing and Evaluation of New SystemAfter the protocol CMDS solution was developedand deployed, CMR technical and clinical/researchteam members engaged in a multi-month iterativetesting and evaluation period. Testing andevaluation were conducted by the CMRclinical/research team through the use of virtualmachines equipped with Windows 7 and populatedwith extensive dummy data. The CMR technicalteam modified the system in response to thefeedback they received from the CMRclinical/research team. A second iterative testingand evaluation period followed deployment ofCMDS v1.0, during which time participatingradiology practices were asked to pilot test CMDSv1.0 alongside the existing Access system byentering a subset of patient records into CMDSv1.0 and providing feedback to the CMR technicalteam regarding features in need of modification.The CMR technical team modified the systemaccordingly before deployment of CMDS v1.1.

The testing period was implemented to ensure theclinical utility and completeness of the datacollection form, letter templates, report templates,and system UI, as well as the accuracy of the data

3

collection process and the overall functionality ofthe new system.

III. ResultsTechnical Functionality of the CMDSOverview of Process and Database SchemaFigure 1 provides an overview of the CMR datacapture and storage process; Appendix A providesa partial schema of the CMDS, showing patient-relevant tables. Currently, eight radiology practicesuse the CMDS. The system is designed such thateach radiology practice maintains an on-site SQLdatabase. Users interact with the system via anintranet-based CMDS UI running on Windows 7+(Appendix B). Data obtained during patient visitsinclude risk factors, reason for and type of imagingperformed, radiology findings and interpretation,recommendations for follow-up, and pathologyresults (if performed). The data are manuallyentered into the CMDS through one of twomethods: direct data entry or data entry from astandardized, bilingual (English/Spanish) datacollection form (Appendix C).

Data from patients who agree to participate in theCMR for research are exported from the localCMDS SQL database to the central CMR SQLdatabase located at and maintained by the Officeof Information Systems in the School of Medicineat our institution (Figure 1). Data transfer occurson a biannual or annual basis. Each radiologypractice conducts a query of the local SQL server,and the data are captured and encoded by site andpatient medical record number (MRN). The dataare then masked, zipped, and auto-encrypted andtransferred by SFTP to the CMR team for importinto the central CMR SQL database.

Security FeaturesStringent security measures include the use ofdummy data for testing at the central CMR site,password protection for entry into the CMDS UI atparticipating radiology practices and access to thecentral CMR database, masking of identifiersbefore encryption, and the use of GPG4win forencryption. In addition, a single delegate at eachradiology practice has access to the encryption keybut not the decryption key; only the central CMRadministrator has access to both the encryptionand decryption keys and hence, access to fullyidentified data for research purposes.

Technical Fixes Post DeploymentThe CMDS has been in continuous developmentsince deployment. The majority of fixes occurredafter deployment of the protocol CMDS, with littleongoing need for fixes since “live” deployment ofCMDS v1.1. Most fixes have been minor and inresponse to requests of the radiology practices orthe CMR clinical/research team. For example, acommon software issue has been the incorrectspelling of a clinical term or a user request tochange the spelling of a term to the preferred (orupdated) clinical or research usage. Other softwareissues have been more impactful, yet easy to fix.For instance, the initial system design did not takeinto account the fact that women can have normalradiology findings for one breast and abnormalfindings for the other breast. Before the fix wasapplied to delineate separate fields for the left andright breast, two separate follow-up letters were

4

Figure 1. An overview of the CMR and CMDSsystem.

generated, one indicating that the radiologyfindings were normal, and the other indicating thatthe findings were abnormal. Additionally, severalradiology practices requested additionaldocumentation on the dates when letters wereprinted and mailed to each patient. Specifically,the initial system would track only the print dateof the most recent letter of a specific type for anindividual patient; previous print dates for thesame type of letter were overwritten by the system.To address this, changes were made to portions ofthe database structure related to letters, such thatthe system now tracks the print date of each letterby type of letter and by patient. The system alsonow has the capability to generate a list of allprinted letters stamped by date for each patient,for a subgroup of patients, or for a type of letterwithin a specified timeframe. Another softwareissue involved changes in patient mailing address.Initially, when a patient had a change in mailingaddress, radiology practices were able to updatethe address in the system, but they were unable toprint a second copy of a letter, such as the ”VisitReminder” letter, to mail to the new address. Inthis case, the software fix was to incorporate afeature to allow sites to manually print secondcopies of letters for individual patients.

Rather substantial changes were made to thelayout and configuration of the UI, although thesewere mostly stylistic and involved therearrangement of menus on various screen displays.Several dropdown menus also were expanded toincrease the granularity of the data that arecaptured.

IV. Application of the System Regulatory ComplianceOne of the primary applications of the CMDS is tomeet the patient follow-up and regulatoryreporting requirements set forth by MQSA andBDMRA. The CMDS accomplishes this largelythrough the automated generation of patientfollow-up letters and regulatory reports. Twenty-two letter templates have been created and areclassified into 3 categories: follow-up/appointmentreminders; mammography and pathology reports;and non-mammography imaging reports (i.e.ultrasound, MRI, CT). The letters include follow-

up reminders for upcoming appointments,regulation-compliant reports of mammography andother imaging findings, and clinicalrecommendations (e.g., consultation with yourphysician regarding breast cancer risk screening,screening options/frequency).

Thirty-nine report templates have been createdand are classified into 5 categories: audit reports;tracking reports; patient lists; mammography andpathology reports; and other reports. The reportsinclude required auditing reports (e.g., MQSAtracking requirements), lists of all patients whoreceived appointment reminder letters, reports ofall patients who received specific mammographyresults (e.g., positive, outstanding [positive but notyet followed-up], negative, indeterminate, etc.), andreports of all patients who had biopsies performedby pathology.

Of note, the letter and report templates can betailored to meet the needs and desires of a givenradiology practice. This is accomplished via theCMDS UI, without the need for additional softwareprogramming by the central CMR administrator.

ResearchIn addition to providing automated letters andreports, another primary application of the CMDSis to facilitate research. This is accomplishedprimarily through an export functionality thatallows an auto-generated “dump” of patient,radiology, and pathology data from each radiologypractice’s local SQL database. The data are sent tothe CMR team and loaded into the central CMRSQL server for processing and long-term storage.The data are used for population-level research onbreast imaging and clinical outcomes. The CMDSsupports the CMR and several other federallyfunded research projects, including a new projectfunded by the National Cancer Institute that aimsto adapt the CMDS to support a Lung CancerScreening Registry.

V. DiscussionPrincipal FindingsThe CMDS: (1) assists participating CMRradiology practices in meeting the reportingrequirements of MQSA and BDMRA through the

5

automated generation of follow-up letters andreports; (2) provides a data repository for radiologypractices and a data collection tool for researchersinterested in population-level research on breastimaging and clinical outcomes; and (3) improvespatient care by facilitating the tracking and follow-up of patients and by promoting innovativeoutcomes-based research.

We note a few interesting features of the CMDSsystem. First, the majority of participating CMRradiology practices are small practices located inprimarily rural counties in North Carolina. Theseradiology practices generally do not have access toan MIS or mammography-supported EMR system,which is typical for rural areas of the United States[5,7]. As such, our CMDS system incorporates acombination of paper data collection, manual dataabstraction from radiology and pathology reports,and manual data entry via the intranet-basedCMDS UI. The data collection form isstandardized, bilingual, and requires minimal effortto complete, with one page of patient-reported dataand one page of radiologist-reported data. TheCMDS UI likewise requires minimal effort ortraining in data entry.

While the United States government has advocatedfor universal adoption of EMR systems byhealthcare providers, offering incentives such as the“Meaningful Use” program enacted as part of theHealth Information Technology for Economic andClinical Health Act, providers have been slow toadopt EMR systems, in part because of costs, realor perceived reductions in provider productivity,and lack of incentives for sharing patient data [8].This is especially true in rural areas, despite theadditional incentives provided by the RegionalExtension Center program [7]. Moreover, evenamong institutions that have an EMR system inplace, EMR data are rarely incorporated intoradiology workflows or Radiology InformationSystems [9], which was the situation for themajority of our participating CMR radiologyclinics. Thus, the custom CMDS provides astreamlined solution to assist with the patientfollow-up and reporting requirements of MQSA andBDMRA among radiology practices without access

to an MIS or mammography-supported EMRsystem.

Security features of the CMDS also are vital to itssuccess. These security features include passwordprotection for authentication and access andGPG4Win encryption of data at participatingradiology practices. For data capture and storagewithin the central CMR database, a manualmethod was chosen, in which participatingradiology practices capture and encode the data bysite and patient MRN stamps, mask all identifiersbefore encryption, zip and auto-encrypt the data,and transfer the data via SFTP to the central site.The encryption key is maintained by a singledelegate at each site and by the central CMRadministrator, but only the central administratorholds the decryption key, which opens access tofully identified data for research purposes. Incombination, these features help to ensure thatidentified patient data stored in the CMDS aresecure at all stages of the data collection, entry,and export process.

LimitationsThe CMDS was developed specifically for the CMRand to assist participating radiology practices withcompliance with MQSA and BDMRArequirements. A limitation is that the systemrequires continuous updating to meet evolvingfederal regulations. However, the system wasdesigned with this in mind, and mostadministrative tasks are streamlined and requirelittle programming by the central CMRadministrator. Another limitation is thegeneralizability of the system. While the generalapproach and system design of the system can bemodified for other applications and for otherinstitutions, modification requires a certain level oftechnical and clinical expertise.

ConclusionsThe CMR was founded in 1994 and supportedinitially by an Access-based data system. The new,custom, SQL-based CMDS v1.1 contains numerousfeatures and capabilities that were not present inthe earlier data system, including a user-friendlyUI, an unlimited number of data fields,compatibility with the latest versions of Windows,

6

and multi-user operability. Importantly, the CMDSprovides an enterprise-level system that assistsparticipating CMR radiology practices withcompliance with the regulatory requirements setforth in MQSA and BDMRA. The CMDS alsoserves as a data repository for participatingradiology practices and a data collection tool forresearchers interested in population-level researchon breast imaging and clinical outcomes. Thesystem currently supports several federally fundedresearch projects in addition to the CMR. Finally,the CMDS improves patient care by facilitating thetracking and follow-up of patients and by fosteringinnovative clinical research.

AcknowledgmentsThis project was supported by the RenaissanceComputing Institute and an award from theNational Cancer Institute (P01CA154292) to Dr.Diana Miglioretti of the University of California atDavis and Dr. Karla Kerlikowske of the Universityof California at San Francisco.

The authors thank the participating women,mammography facilities, and radiologists for theirsupport of the Carolina Mammography Registry.

The authors also wish to acknowledge KaramarieFecho for assistance with the writing and editing ofthe technical report, and Karen Green and PatrickSchmitt for assistance with the publication design.

How to cite this paperKapeljushnik, O., Schmitt, C., Henderson, L. M., &Greenwood-Hickman, M. A. (2017): CMDS: A Data System for a Multi-institutional Mammography Registry. RENCI, University of North Carolina at Chapel Hill. Text. https://doi.org/10.7921/G0639MW0

About the AuthorsOleg Kapeljushnik is a Software Developer atRENCI and led the development of the CMDS.Charles Schmitt formerly served as ChiefTechnology Officer and Director of Informatics atRENCI and now serves as Director of Data Scienceat the National Institute of Environmental HealthSciences. Kapeljushnik designed, implemented, andtested the CMDS under the scientific direction ofSchmitt. Louise Henderson is an AssociateProfessor within the Department of Radiology atUNC and Director of the CMR. Hendersonprovided the conceptual vision for the CMDS.Anne Greenwood-Hickman formerly served asData Manager for the CMR.

7

References

1. Grady D. American Cancer Society, in a shift, recommends fewer mammograms. The New York Times. October 20, 2015. http://www.nytimes.com/2015/10/21/health/breast-cancer-screening-guidelines.html?_r=0. Archived at: http://www.webcitation.org/6d769ubyv

2. Oeffinger KC, Fontham ETH, Etzioni R, et al. Breast cancer screening for women at average risk.2015 guideline update from the American Cancer Society. JAMA 2015;314(15):1599-614. PMID: 26501536

3. Ho JM, Jafferjee N, Covarrubias GM, et al. Dense breasts: a review of reporting legislation and available supplemental screening options. AJR Am J Roentgenol 2014;203(2):449-56. PMID: 25055284

4. Schattner E. Mandatory reporting of breast density in screening mammograms. Why the controversy? Forbes January 29, 2015. http://www.forbes.com/sites/elaineschattner/2015/01/29/mandatory-reporting-of-breast-density-in-screening-mammograms-why-the-controversy/. Archived at: http://www.webcitation.org/6d76TN0bM

5. Slight SP, Berner ES, Galanter W, et al. Meaningful use of electronic health records: experiences from the field and future opportunities. JMIR Med Inform 2015;3(3):e30. PMID: 26385598

6. Hardy K. Mammography information systems. Radiology Today 2012;13(11):28.7. Samuel CA. Area-level factors associated with electronic health record adoption and meaningful

use in the regional extension center program. J Am Med Inform Assoc 2014;21:976–83. PMID: 24798687

8. Mennemeyer ST, Menachemi N, Rahurkar S, et al. Impact of the HITECH act on physicians’ adoption of electronic health records. J Am Med Inform Assoc 2016;23(2):375–9. PMID: 26228764

9. Khorasani R. Can you efficiently incorporate patient-specific electronic medical record data into radiology workflow? J Am Coll Radiol 2012;9(12):862-3. PMID: 23206640

8

Appendix A: Partial schema of the CMDS, showing patient-relevant tables.

9

Appendix B: Screenshots of the CMDS UI.A.

B.

10

Appendix C: Data collection form used to capture patient and radiology data formanual entry into the CMDS.

A. Page 1: Patient form.

11

B. Page 2: Radiology form.

12